Present disclosure provide a linear prediction-based noise signal processing method includes: acquiring a noise signal, and obtaining a linear prediction coefficient according to the noise signal; filtering the noise signal according to the linear prediction coefficient, to obtain a linear prediction residual signal; obtaining a spectral envelope of the linear prediction residual signal according to the linear prediction residual signal; and encoding the spectral envelope of the linear prediction residual signal. According to the noise processing method, the noise generation method, the encoder, the decoder, and the encoding and decoding system that are in the embodiments of the present disclosure, more spectral details of an original background noise signal can be recovered, so that comfort noise can be closer to original background noise in terms of subjective auditory perception of a user, and subjective perception quality of the user is improved.
Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A noise signal processing method, comprising: obtaining, by an encoder comprising a processor, a linear prediction coefficient according to an acquired noise signal corresponding to an audio signal; filtering, by the encoder, the noise signal according to the linear prediction coefficient to obtain a linear prediction residual signal; obtaining, by the encoder, a spectral envelope of the linear prediction residual signal according to the linear prediction residual signal; obtaining a spectral detail of the linear prediction residual signal according to the spectral envelope of the linear prediction residual signal; quantizing, by the encoder, the spectral envelope of the linear prediction residual signal; quantizing, by the encoder, the spectral detail of the linear prediction residual signal; and writing, by the encoder, the quantized spectral envelope of the linear prediction residual signal and the quantized spectral detail of the linear prediction residual signal into a bitstream for storing or transmitting.
An audio encoder processes noise signals by first calculating a linear prediction coefficient from the noise. This coefficient is then used to filter the noise signal, producing a linear prediction residual signal. A spectral envelope and spectral detail of this residual signal are then calculated. Both the spectral envelope and the spectral detail are quantized before being written into a bitstream. This bitstream is then available for storage or transmission.
2. The noise signal processing method according to claim 1 , wherein after the filtering the noise signal according to the linear prediction coefficient to obtain a linear prediction residual signal, the method further comprises: obtaining energy of the linear prediction residual signal according to the linear prediction residual signal, and wherein the quantizing the spectral detail of the linear prediction residual signal comprises: quantizing the linear prediction coefficient, the energy of the linear prediction residual signal, and the spectral detail of the linear prediction residual signal.
The noise signal processing method of claim 1 involves calculating the energy of the linear prediction residual signal after filtering the noise signal. The quantization process includes quantizing the linear prediction coefficient, the energy of the linear prediction residual signal, and the spectral detail of the linear prediction residual signal. So, a linear prediction coefficient is calculated from the noise, which is used to filter the noise and get the linear prediction residual signal. Then calculate the energy of this signal. Finally, quantize the linear prediction coefficient, the energy, and spectral detail and output to a bitstream.
3. The noise signal processing method according to claim 2 , wherein the obtaining a spectral detail of the linear prediction residual signal according to the spectral envelope of the linear prediction residual signal comprises: obtaining a random noise excitation signal according to the energy of the linear prediction residual signal; and using a difference between the spectral envelope of the linear prediction residual signal and a spectral envelope of the random noise excitation signal as the spectral detail of the linear prediction residual signal.
In the noise signal processing method described in claim 2, the spectral detail of the linear prediction residual signal is obtained by first generating a random noise excitation signal based on the energy of the linear prediction residual signal. The difference between the spectral envelope of the linear prediction residual signal and the spectral envelope of the random noise excitation signal is then used as the spectral detail. So, in addition to the steps of claim 2, a random noise excitation signal is generated from the residual signal's energy and its spectral envelope is compared to the residual signal's spectral envelope; the difference is the spectral detail used in the quantization.
4. The noise signal processing method according to claim 1 , wherein the obtaining the spectral detail of the linear prediction residual signal according to the spectral envelope of the linear prediction residual signal comprises: obtaining a spectral envelope of first bandwidth according to the linear prediction residual signal, wherein the first bandwidth is within a bandwidth range of the linear prediction residual signal; and obtaining the spectral detail of the linear prediction residual signal according to the spectral envelope of the first bandwidth.
In the noise signal processing method of claim 1, the spectral detail of the linear prediction residual signal is obtained by first determining a spectral envelope of a first bandwidth within the total bandwidth of the linear prediction residual signal. The spectral detail is then derived from this narrower bandwidth spectral envelope. So, instead of processing the spectral envelope of the whole linear prediction residual signal for the spectral detail, only a portion is used.
5. The noise signal processing method according to claim 4 , wherein the obtaining the spectral envelope of first bandwidth according to the spectral envelope of the linear prediction residual signal comprises: calculating a spectral structure of the linear prediction residual signal; and using a spectrum of a first part of the linear prediction residual signal as the spectral envelope of the first bandwidth, wherein a spectral structure of the first part is stronger than a spectral structure of the remaining part of the linear prediction residual signal.
In the noise signal processing method of claim 4, a spectral structure of the linear prediction residual signal is calculated. The spectrum of a "first part" of the linear prediction residual signal is used as the spectral envelope of the defined "first bandwidth", where the "first part" exhibits a stronger spectral structure than the remaining portion of the residual signal. So, calculate the spectral structure of the linear prediction residual signal and then use the part of that spectrum with the strongest spectral structure as the spectral envelope.
6. The noise signal processing method according to claim 5 , wherein the spectral structure of the linear prediction residual signal is calculated in one of the following manners: calculating the spectral structure of the linear prediction residual signal according to a spectral envelope of the noise signal; and calculating the spectral structure of the linear prediction residual signal according to the spectral envelope of the linear prediction residual signal.
In the noise signal processing method of claim 5, the spectral structure of the linear prediction residual signal can be calculated using either the spectral envelope of the original noise signal, or the spectral envelope of the linear prediction residual signal itself. So, in addition to the steps of claim 5, the calculation of the spectral structure can be done two different ways; one using the original noise signal's spectral envelope and the other using the linear prediction residual signal's spectral envelope.
7. The noise signal processing method according to claim 1 , wherein after the obtaining the spectral detail of the linear prediction residual signal according to the spectral envelope of the linear prediction residual signal, the method further comprises: calculating a spectral structure of the linear prediction residual signal according to the spectral detail of the linear prediction residual signal; and obtaining a spectral detail of second bandwidth of the linear prediction residual signal according to the spectral structure, wherein the second bandwidth is within a bandwidth range of the linear prediction residual signal, and a spectral structure of the second bandwidth is stronger than a spectral structure of the remaining part of bandwidth of the linear prediction residual signal, and wherein the quantizing the spectral envelope of the linear prediction residual signal comprises: quantizing the spectral detail of the second bandwidth of the linear prediction residual signal.
In the noise signal processing method of claim 1, after the spectral detail is obtained, a spectral structure of the linear prediction residual signal is calculated based on the spectral detail. A spectral detail of a "second bandwidth" (within the overall bandwidth of the residual signal, and having a stronger spectral structure than the remaining bandwidth) is then obtained, and the quantization process includes quantizing this "second bandwidth" spectral detail of the linear prediction residual signal.
8. A comfort noise signal generation method, comprising: decoding, by a decoder comprising a processor, a received bitstream to obtain a spectral detail and a linear prediction coefficient, wherein the spectral detail indicates a spectral envelope of a linear prediction excitation signal, and wherein the spectral detail is obtained and quantized by an encoder according to a spectral envelope of a linear prediction residual signal; obtaining, by the decoder, the linear prediction excitation signal according to the spectral detail; obtaining, by the decoder, a comfort noise signal according to the linear prediction coefficient and the linear prediction excitation signal; and reconstructing, by the decoder, an audio signal according to the obtained comfort noise signal.
A method for generating comfort noise involves decoding a received bitstream to obtain a spectral detail and a linear prediction coefficient. The spectral detail, which represents a spectral envelope of a linear prediction excitation signal, was previously obtained and quantized by an encoder based on a spectral envelope of a linear prediction residual signal. Using the decoded spectral detail, a linear prediction excitation signal is generated. This signal, along with the linear prediction coefficient, is then used to create a comfort noise signal, which is used to reconstruct an audio signal.
9. The comfort noise signal generation method according to claim 8 , wherein the spectral detail is the spectral envelope of the linear prediction excitation signal.
In the comfort noise signal generation method of claim 8, the spectral detail directly represents the spectral envelope of the linear prediction excitation signal. So, the decoded spectral detail represents the spectral envelope of the linear prediction excitation signal and is used to create the comfort noise.
10. The comfort noise signal generation method according to claim 8 , wherein the bitstream comprises energy of linear prediction excitation, and before the obtaining the comfort noise signal according to the linear prediction coefficient and the linear prediction excitation signal, the method further comprises: obtaining a first noise excitation signal according to the energy of the linear prediction excitation; and obtaining a second noise excitation signal according to the first noise excitation signal and the linear prediction excitation signal, and wherein the obtaining the comfort noise signal according to the linear prediction coefficient and the linear prediction excitation signal comprises: obtaining the comfort noise signal according to the linear prediction coefficient and the second noise excitation signal.
This invention relates to generating comfort noise signals in audio processing, particularly for maintaining natural-sounding background noise during speech or audio transmission interruptions. The problem addressed is ensuring smooth transitions and natural-sounding noise when audio signals are interrupted, such as in voice-over-IP or telecommunication systems. The method involves generating a comfort noise signal using linear prediction (LP) techniques. The bitstream contains energy information of the linear prediction excitation signal. Before generating the comfort noise signal, the method extracts a first noise excitation signal based on this energy. A second noise excitation signal is then derived by combining the first noise excitation signal with the linear prediction excitation signal. The comfort noise signal is finally generated using the linear prediction coefficients and this second noise excitation signal. This approach ensures that the comfort noise maintains the spectral characteristics of the original signal while smoothly filling gaps in audio transmission, improving user experience by avoiding abrupt silence or unnatural noise artifacts. The method is particularly useful in real-time communication systems where maintaining perceptual continuity is critical.
11. An encoder, comprising: a memory for storing processor-executable instructions; and a processor operatively coupled to the memory, the processor being configured to execute the processor-executable instructions to: obtain a linear prediction coefficient according to an acquired noise signal; filter the noise signal according to the obtained linear prediction coefficient, to obtain a linear prediction residual signal; obtain a spectral envelope of the linear prediction residual signal according to the linear prediction residual signal; quantize the spectral envelope of the linear prediction residual signal; obtain a spectral detail of the linear prediction residual signal according to the spectral envelope of the linear prediction residual signal; quantize the spectral detail of the linear prediction residual signal; and write the quantized spectral envelope of the linear prediction residual signal and the quantized spectral detail of the linear prediction residual signal into a bitstream for storing or transmitting.
An encoder includes memory storing instructions and a processor. The processor is configured to obtain a linear prediction coefficient from an acquired noise signal, filter the noise signal to produce a linear prediction residual signal, and obtain the spectral envelope of that residual signal. The spectral envelope is quantized. The processor also obtains a spectral detail of the residual signal based on the spectral envelope and quantizes that as well. Finally, the quantized spectral envelope and detail are written into a bitstream.
12. The encoder according to claim 11 , wherein the processor is further configured to execute the processor-executable instructions to: obtain energy of the linear prediction residual signal according to the linear prediction residual signal; and quantize the linear prediction coefficient, the energy of the linear prediction residual signal, and the spectral detail of the linear prediction residual signal.
The encoder described in claim 11 is further configured to calculate the energy of the linear prediction residual signal and to quantize the linear prediction coefficient, the energy of the linear prediction residual signal, and the spectral detail of the residual signal. So in addition to the steps of claim 11, the encoder calculates the energy of the residual signal and quantizes it along with the linear prediction coefficient and spectral detail, then writes these to the bitstream.
13. The encoder according to claim 12 , wherein the processor is further configured to execute the processor-executable instructions to: obtain a random noise excitation signal according to the energy of the linear prediction residual signal; and use a difference between the spectral envelope of the linear prediction residual signal and a spectral envelope of the random noise excitation signal as the spectral detail of the linear prediction residual signal.
The encoder described in claim 12 is further configured to generate a random noise excitation signal based on the energy of the linear prediction residual signal. The spectral detail is calculated as the difference between the spectral envelope of the residual signal and the spectral envelope of this random noise excitation signal. So in addition to the steps of claim 12, a random noise excitation signal is generated from the energy of the residual signal, its spectral envelope is determined, and the difference between it and the residual signal's spectral envelope is used as the spectral detail.
14. The encoder according to claim 11 , wherein the processor is further configured to execute the processor-executable instructions to: obtain a spectral envelope of first bandwidth according to the linear prediction residual signal, wherein the first bandwidth is within a bandwidth range of the linear prediction residual signal; and obtain the spectral detail of the linear prediction residual signal according to the spectral envelope of the first bandwidth.
The encoder of claim 11 is further configured to obtain a spectral envelope of a "first bandwidth" portion of the linear prediction residual signal (where the first bandwidth is within the total bandwidth of the signal) and to derive the spectral detail of the residual signal based on this "first bandwidth" spectral envelope. So, instead of using the full signal's spectral envelope, a smaller bandwidth is used to get the spectral detail.
15. The encoder according to claim 14 , wherein the processor is further configured to execute the processor-executable instructions to: calculate a spectral structure of the linear prediction residual signal, and use a spectrum of a first part of the linear prediction residual signal as the spectral envelope of the first bandwidth, wherein a spectral structure of the first part is stronger than a spectral structure of the remaining part of the linear prediction residual signal.
In the encoder of claim 14, the processor calculates a spectral structure of the linear prediction residual signal and uses the spectrum of a "first part" of the residual signal as the spectral envelope of the "first bandwidth", where the "first part" has a stronger spectral structure than the remainder of the signal. So, the encoder is configured to determine the spectrum of the residual signal, determine its spectral structure, and then take the section of the residual signal with the strongest spectral structure.
16. The encoder according to claim 15 , wherein the processor is further configured to execute the processor-executable instructions to: calculate the spectral structure of the linear prediction residual signal according to a spectral envelope of the noise signal; or calculate the spectral structure of the linear prediction residual signal according to the spectral envelope of the linear prediction residual signal.
In the encoder of claim 15, the calculation of the spectral structure of the linear prediction residual signal is performed by calculating the spectral structure either from the spectral envelope of the *original* noise signal, or from the spectral envelope of the linear prediction residual signal itself. So, the encoder is configured to calculate the spectral structure in two ways; once with the original noise signal and once using the residual signal.
17. The encoder according to claim 11 , wherein the processor is further configured to execute the processor-executable instructions to: obtain the spectral detail of the linear prediction residual signal according to the spectral envelope of the linear prediction residual signal, calculate a spectral structure of the linear prediction residual signal according to the spectral detail of the linear prediction residual signal, and obtain a spectral detail of second bandwidth of the linear prediction residual signal according to the spectral structure, wherein the second bandwidth is within a bandwidth range of the linear prediction residual signal, and a spectral structure of the second bandwidth is stronger than a spectral structure of the remaining part of bandwidth of the linear prediction residual signal; and quantize the spectral detail of the second bandwidth of the linear prediction residual signal.
In the encoder of claim 11, the processor is configured to calculate a spectral structure of the linear prediction residual signal based on the spectral detail of the signal. A spectral detail of a "second bandwidth" of the residual signal is then obtained based on this spectral structure, where the "second bandwidth" has a stronger spectral structure than the remaining bandwidth. The encoder then quantizes the "second bandwidth" spectral detail.
18. A decoder, wherein the decoder comprises: a memory for storing processor-executable instructions; and a processor operatively coupled to the memory, the processor being configured to execute the processor-executable instructions to: decode a received bitstream to obtain a spectral detail and a linear prediction coefficient, wherein the spectral detail indicates a spectral envelope of a linear prediction excitation signal, wherein the spectral detail is obtained and quantized by an encoder according to a spectral envelope of a linear prediction residual signal; obtain the linear prediction excitation signal according to the spectral detail; obtain a comfort noise signal according to the linear prediction coefficient and the linear prediction excitation signal; and reconstruct an audio signal according to the obtained comfort noise signal.
A decoder has memory and a processor. The processor is configured to decode a received bitstream to obtain a spectral detail and a linear prediction coefficient. This spectral detail represents a spectral envelope of a linear prediction excitation signal, which was previously obtained and quantized by an encoder based on the spectral envelope of a linear prediction residual signal. The processor then generates the linear prediction excitation signal from the decoded spectral detail, generates a comfort noise signal using the linear prediction coefficient and the excitation signal, and reconstructs an audio signal from the comfort noise.
19. The decoder according to claim 18 , wherein the spectral detail is the spectral envelope of the linear prediction excitation signal.
In the decoder of claim 18, the spectral detail obtained from the bitstream *is* the spectral envelope of the linear prediction excitation signal. So, the spectral detail directly indicates the linear prediction excitation signal.
20. The decoder according to claim 18 , wherein the bitstream comprises energy of linear prediction excitation, and wherein the processor is further configured to execute the processor-executable instructions to: obtain a first noise excitation signal according to the energy of the linear prediction excitation; obtain a second noise excitation signal according to the first noise excitation signal and the linear prediction excitation signal; and obtain the comfort noise signal according to the linear prediction coefficient and the second noise excitation signal.
In the decoder of claim 18, the received bitstream includes the energy of the linear prediction excitation. The processor is further configured to generate a "first noise excitation signal" based on this energy, then generate a "second noise excitation signal" based on the first excitation signal and the linear prediction excitation signal. The comfort noise signal is then generated using the linear prediction coefficient and the *second* noise excitation signal.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
September 29, 2016
August 8, 2017
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.